Summary

  • A drafting project titled the Governance Document for the Recognition, Operation, and Derecognition of Regional Internet Registries is meant to replace ICP-2, the 2001 text whose own title addressed only the establishment of new registries. Its second public consultation concluded on 7 November.
  • Expanding a title is not the same as defining a power. Whether the successor becomes accountable government or a discretionary gate turns on five things the obtained record does not contain: the trigger, the evidence, the decider, the review, and the exit.

The word that changed

Two document titles, twenty-five years apart, and one word that does a great deal of work.

The record published under ICANN's name is titled ICP-2: Criteria for Establishment of New Regional Internet Registries for Internet Number Resources, dated 4 June 2001. Establishment. That is the whole job described in the title: the conditions under which a new regional registry may come into being.

The project that is meant to replace it carries a longer name. According to the account I obtained, "the second public consultation of the Governance Document for the Recognition, Operation, and Derecognition of Regional Internet Registries (RIR Governance Document) concluded on 7 November. This document is intended to replace Internet Coordination Policy 2 (ICP-2), the foundational document developed in 2001 for the recognition and evaluation of new Regional Internet Registries (RIRs)" (CircleID).

Recognition, operation, and derecognition. The third noun is new, and it is the one that matters to anyone whose address space sits inside a registry's books.

I want to be exact about what that sentence establishes and what it does not. It establishes that a document with that title exists as a drafting project and that a consultation stage closed on 7 November. It does not tell us what derecognition means in the text, who may invoke it, what must be proved before it can be invoked, or what happens to a region's registration records afterwards. On the record I could obtain, the drafting project is titled for an exit power; the terms of that exit remain unpublished in the material available to me.

What the record actually says

Strip out the inferential layer and the evidence base here is thin, specific, and worth stating in full so the reader can see exactly where the floor is.

First, the replacement project and its consultation close, quoted above, together with the claim that it would displace the 2001 text.

Second, the identity of the 2001 baseline: its title and its date. I did not inspect its body. Nothing in the record I hold establishes who drafted it, who commissioned it, how it was adopted, or what it says about failure, discipline or exit. Anyone who tells you the 2001 text authorises the removal of a registry is telling you something this record cannot confirm.

Third, the shape of the system being governed. The Number Resource Organization states that "there are five Regional Internet Registries (RIRs) in the world" and that "RIRs manage, distribute, and register Internet number resources (IPv4 and IPv6 address space and Autonomous System (AS) Numbers) within their respective regions" (NRO). Five entities, five regions, one registration function each. A governance document with an exit clause is therefore a document about five single points of institutional failure.

Fourth, and reported rather than verified: that the drafting was led by the NRO Number Council at the request of the NRO Executive Council, that the project envisages requirements for financial and operational independence, dispute resolution and periodic audits, that it includes provisions for non-compliance remediation, and that it contemplates guidelines for emergency operators who would keep critical registry services running during a transition or where a registry cannot adequately perform its functions, with completion planned for 2026. Those are the features attributed to the project in the account I obtained.

They are not clause readings. I have not seen the clause text, the section numbers, the definitions, or a vote.

Fifth, the political weather. One news report ties momentum for the reform to the AFRINIC situation and to demands for legal safeguards and resilient governance models (BusinessWorld). That is that outlet's framing, attributed, not an established causal finding, and no court or institutional record linking a dispute to a drafting line sits in my evidence.

The BTW directory tracks the instrument itself: Governance Document for the Recognition, Operation, and Derecognition of Regional Internet Registries Only for Strategic Circle.

Five questions an exit clause has to answer

A registry is not a club that can quietly stop meeting. It holds a regional allocation record on which operators, members and transferees depend. So the interesting question is not whether the drafting project mentions derecognition — its title already does — but whether the resulting text would survive the only test that matters: can a finding against a registry be made, contested and enforced without the deciding institution marking its own homework?

There are five places to look, and each one is a place where the obtained record is silent.

The trigger. What counts as failure? A closed list of objective conditions an auditor can check, or a judgement call about whether a registry is performing "adequately"? The difference is not semantic. An objective trigger lets an operator know in advance what conduct puts its recognition at risk. A discretionary trigger hands the threshold to whoever holds the pen. Reported summaries mention non-compliance remediation, which suggests a trigger exists in the draft. Whether it is written as a checklist or a standard, I cannot say from this record.

The evidence. Who proves failure, and to what standard? A financial audit, a service-continuity test, a member complaint, and an institutional impression are four different evidentiary regimes with four different error rates. If the draft ties a finding to audited accounts or measured service levels, the power to derecognise is bounded by numbers. If it ties a finding to an assessment of independence and resilience, the power is bounded by judgement. The reported mention of periodic audits and independence requirements points toward the first. The clause text would settle it.

The decider. This is where provenance stops being a technicality. The drafting is reported to run from an instruction by the NRO Executive Council to the NRO Number Council. That is an internal chain within the family of institutions whose members would be subject to the instrument. Internal authorship is not misconduct — self-regulatory bodies write their own standards all the time, and they often write better ones than outsiders would. But it does raise a specific design question: is the body that finds failure separate from the body that benefits from the finding? A framework can be internally drafted and still be credible, provided the adjudicating function sits somewhere the reviewed registry does not control.

The review. What can a registry do when a finding lands? Produce records, answer the case, seek a second opinion, appeal to the other registries, or go to an external forum? Response rights are not decoration. They are the mechanism that converts a finding from an announcement into a decision with a record. The obtained material does not tell me whether the draft contains a response stage, a hearing, an appeal route, or any external check at all. Non-observation is not absence — it is simply the edge of what I hold.

The exit. What happens to the region? If recognition ends, do the registration records travel with a successor, does an emergency operator step in, are members' allocations preserved, and who answers for transfers already in flight? The reported notion of emergency operators keeping critical services running suggests the drafters have thought about continuity. A title that adds derecognition to a document that also has to describe what survives derecognition is doing something genuinely difficult. Whether it does that thing well is exactly what cannot be assessed from a summary of the features.

Provenance is a question, not a verdict

There is a lazy version of this argument and a serious one.

The lazy version says the registries are writing the rule that judges the registries, therefore the rule is captured. That reading is unsupported here on two counts. It assumes a commissioning history for the 2001 text that my record does not establish, and it treats "review" as adjudication when the word may equally describe periodic evaluation, reporting or remediation. It also converts an institutional location into a finding about motive, which is a jump no evidence in this record makes for me.

The serious version says something narrower and harder to dismiss. A recognition instrument that acquires an exit power becomes a gate. Gates have to be designed, and design choices about who triggers, who proves, who decides, who reviews and what survives are exactly the choices that determine whether a gate disciplines conduct or merely distributes leverage. Internal provenance does not decide that question. It only tells you where to look for the answer, and it tells you that the answer will be written by people with a stake in it.

That is why the title change deserves more attention than a title normally gets. Naming a document for derecognition is a public commitment of scope. It tells every registry in the system, and every operator who depends on one, that the question of how a registry's standing ends is now on the table. Once that commitment is made in a title, the drafting has to answer for it. A document that names an exit and then leaves the terms vague is worse than one that never named it, because the name travels further than the footnote.

Why an operator should read this

The five registries are not the only parties with something at stake. Every network that holds space under a registry's record has an interest in that registry continuing to exist, or in a successor continuing to answer for the same records.

Consider what a derecognition actually touches. Registration data. Reverse DNS delegations. Transfer requests mid-flight. The expectation, held by members and by counterparties, that the entity which issued an allocation can confirm it. None of that is abstract, and none of it is fixed by a title.

If the successor document establishes a credible exit — objective triggers, evidence standards, a decider at arm's length, response rights, and a continuity path for the records — then adding derecognition to the framework is a strengthening move: it gives the system a remedy it previously lacked, and it gives registries a reason to meet published standards.

If instead the exit sits with the same institutional chain that reviews the registry, with an undefined trigger and no response stage, then the document would hand the recognition layer a discretionary power over five single points of failure in the global registration system. That is not a prediction. It is the branch that remains open while the clauses are not public.

What would change this reading

The reading is falsifiable, which is the point of stating it with the boundary attached.

If the operative text turns out to contain an independent decision-maker, a defined evidentiary standard, a response and appeal stage, and continuity or portability protections for records and allocations, then internal provenance would not defeat the instrument's legitimacy — and this article's scepticism would look overstated in the specific way that good drafting is supposed to defeat.

If the text turns out to define derecognition only as periodic assessment with no failure consequence at all, then the premise that the project supplies a remedy for a failing registry needs withdrawing, and the title is doing marketing work the body does not support.

If the text leaves trigger, proof, decision and resource disposition to incumbent-controlled discretion with no effective review, then the recognition-chokepoint reading stops being a hypothesis and becomes a description — but only after someone reads the clauses.

And if the consultation record shows that all five registries authorised the process and the resulting instrument, the independence question changes shape: a jointly authorised rule is a different object from one institution's instruction.

None of those checks can be run on a title and a summary. They can only be run on the text. Which is why the most useful thing to say about the Governance Document for the Recognition, Operation, and Derecognition of Regional Internet Registries today is also the least satisfying: the exit is now named, the consultation has closed, and the terms of the exit have not been published in the material I could obtain. Until they are, the accountability test is unrun, not passed.

Sources